Highlight Extractor

Highlight-Extractor is a service that extracts highlights by subtitle(scripts), resizing video for short-form(16:9) also adds subtitles to videos using OpenAI and OpenAI's Whisper model for subtitle extraction and FFmpeg for embedding subtitles into the video. The service is implemented with FastAPI to provide a API interface.

Requirements

• Python 3.8 or greater

GPU

Execution requires the following NVIDIA libraries to be installed:

• cuBLAS for CUDA 12
• cuDNN for CUDA 12

Installation

1. Clone the repository.

git clone https://github.com/Dockerel/highlight-extractor.git
cd highlight-extractor

2. Make virtual environment.

python -m venv .venv
source .venv/bin/activate

3. Set .env file.

SMTP_ID={YOUR_GOOGLE_SMTP_ID}
SMTP_PW={YOUR_GOOGLE_SMTP_PW}

OPENAI_API_KEY={YOUR_OPENAI_API_KEY}

AWS_ACCESS_KEY_ID={YOUR_AWS_ACCESS_KEY_ID}
AWS_SECRET_ACCESS_KEY={YOUR_AWS_SECRET_ACCESS_KEY}
AWS_BUCKET_NAME={YOUR_AWS_BUCKET_NAME}

4. Execute with command below

at local

uvicorn app.main:app --host 127.0.0.1 --port 8000

to deploy

uvicorn app.main:app --host 0.0.0.0 --port 8000

API Paths

Path	Description
POST /extract-highlights	Used to allocate a task for highlight extracting.
GET /task-status/{task_id}	Used when checking status of tasks by task id.
GET /select-highlight/{task_id}	Used to select favorite extracted highlight.
GET /clear-status	Used to clear in-memory total process status.